Problem Determination of your WebSphere …...Allow users to create, enable and disable validation...

© 2010 IBM Corporation

WebSphere

Problem Determination of yourWebSphere Applications with RAD

Anita Rass Wan, RAD Product Manager

© 2010 IBM Corporation2

IBM Labs Come To You

Agenda

Governance of Code• Creating your own rules - customizing rule templates• Advanced Rule creation• Development process integration• Best Practices

Code coverage and Unit test optimization• Overview• Development process integration

Profiling

Problem scenarios: What should I use?– Crashes, hangs– Memory leaks– Execution bottlenecks

Tuning the JVM

Tuning WebSphere Applications

Questions



Static Analysis in Rational Application Developer 8.0

Java Code Review to find problems of various types:– Design Principles– Globalization– J2EE & J2SE Best Practices– J2EE & J2SE Security– Naming– Performance– Private API

Rich integrated results view provides click to source navigation– Explanations, examples, and quick fixes for problems

Allow users to create, enable and disable validation rules

Allow users to create their own rules based on rule templates

Produce HTML/PDF reports with violations and metrics

Complete Java Code Review (200+ rules)



Getting Started with Static Analysis

Launched through SoftwareAnalyzer configuration

Can create multiple configurations– Scope – workspace, working set or

selected projects– Select and configure rules – focus

on a particular aspect for analysis



Analysis Configuration – Rules

Rule sets– Grouping of selected and configured rules

that are to be run against the resources inthe scope of the analysis

Custom rule sets– Shared through export and import

Rule configuration– Select which rules should be used in the

current configuration– Configure severity of the violations– Change parameters of rules



Adding Custom Rules

Static analysis component comes with several templates that can be used to define customrules through the preferences



Advanced Rule Creation

Why?– Sometimes the prepackaged or template based rules aren’t specific enough– Domain specific constraints' or company wide governance enforcement rules

How?– Software Analyzer component is extensible– Can create new categories/subcategories for your new rules– API framework makes it easier to implement

• Rules, quickfixes, and even custom reports

What?– Need to have some background in

• Eclipse• Java• Eclipse JDT for Java rules



Example – Rule to check for unused imports

New category to hold our rule

New rule definition as part of the new category

Extend the abstract class AbstractCodeReviewRule– Implement analyze()

Optionally implement an associated quickfix by extending JavaCodeReviewQuickFix– Implement fixCodeReviewResult()



Analyzing Results

Double click andview the violation in

the source view.

Generate reportsHTML/PDF

reports.

Export results inXML for otherreporting tools



Software Analyzer Reports

Create HTML/PDF reports of different metrics– Results, Ignored Results, Severity Summary,

Severity by Category



Best Practices

Run Analysis early and often– As part of developer best practice to run periodically– Cost to fix defects rises the further they get into the application’s lifecycle– Logical separation of your application in terms of groups for analysis– Java code for business logic– Java code for connection services– Java code for presentation

Smaller groups of analysis to avoid overwhelming number of results– Especially when first beginning to use the code review capability– Can increase the scope of analysis as code quality improves

Create multiple rule sets to apply to different group types– Globalization rules apply more for java code involved in presentation– Different requirements for code complexity enforcement for business logic vs presentation tier



Integrating Static Analysis Code Reviews – Builds – Step 1

Export your rule set (*.dat) from the Analyzer Configuration



Integrating Static Analysis Code Reviews – Builds – Step 2

Add a step to your build process to invoke the analysis.

setWORKSPACE="C:\PROGRA~1\IBM\TeamConcertBuild\buildsystem\buildengine\eclipse\buildDir"

set ECLIPSE="C:\PROGRA~1\IBM\SDP_1"

%ECLIPSE%\eclipse.exe-application com.ibm.xtools.analysis.commandline.AnalyzeApplication-rulefile "d:/rules.dat"-data %workspace%-directory %workspace%-exportdirectory %WORKSPACE%\RADcodeReviewXML



Agenda

Static Analysis of Code• Creating your own rules - customizing rule templates• Development process integration• Best Practices

Code Coverage• Overview• Development process integration

Advanced Code Review– Creating your own rules – write your own with the Software Analyzer API

Profiling


JVM tuning


Questions



Code Coverage Provides detailed information on what code paths have been encountered during program execution Powerful tool to help determine xUnit test coverage, potential dead code Command line and Ant capability for build integration

– JUnit, code coverage data collection and html report generation



16

Share code coverage information from automated testcase executionImprove test coverage and quality based on code coverage results

RationalApplicationDeveloper

RationalApplicationDeveloper

RationalTeam Concert

DeveloperRational

Team ConcertDeveloper



Feedback metrics help identify shortfalls in test coverage

Process feedbackThe project lead will see that the code

coverage results are at expected levels andresolves the defect

Work item resolvedDeveloper checks code into a

build and resolves defect whenhe sees the code coverage

results for the build

Problem determinationDeveloper reviews build results andimports the code coverage results

directly into the source code to see whatsource is missed during the test runs

New tests addedNew tests are added to cover the

code that has not been covered byprevious tests. Runs tests locally

to ensure improved coverage

DevelopmentA defect is opened

against the build that hadthe noticable code

coverage level drop

Process feedbackProject lead notices adrop in code coverage

from previous build

RationalRationalApplicationApplicationDeveloperDeveloper

++Rational TeamRational Team

ConcertConcert



Code Coverage Integration with RTC

Build system integration– Installable RTC Build System extension– Creates code coverage statistics for BVT tests– Data grouped with build results

…Build Machine 1

Build Machine 2

Build Engine

Build Engine

Code CoverageExtension

Code CoverageExtension



Code Coverage Integration with RTC

Client side– RTC Build details viewer extension – RAD and Web browser

• Show summary of code coverage statistics• Additional Code Coverage tab to show detailed coverage statistics report

– Integrated work item search and creation– Import to local workspace for rich viewing in navigator and source views

Demo



Agenda



Profiling


Tuning the JVM


Questions



What is Profile on Server?

Profile on Sever is the ability to look at various performance aspects ofabout your Java application

Similar to Run and Debug on Server. Requires an Agent Controller (AC) running on the target machine to

connect to the JVM Users start the profile action from a web resource or a server instance in

Server Views.



Agent Controller

A daemon process that enables client applications to launch host processes and interactwith agents that coexist within host processes.

An Agent Controller is required to be running on the target machine being profiled Integrated Agent Controller (IAC) is embedded in RAD for local use case. No additional installation, Automatically enabled. Start and stop on demand.

Standalone Rational Agent Controller ( RAC) is required for remote profiling. Requires additional installation on target WAS machine.



Launch profile on server



New in RAD v8 – Easy profiling setup

• If using an older RAC on a remote machine, RAD will auto detect theolder version and prompt to add environment variables or upgrade yourRAC



Quick start Wizard



Select Memory Analysis Option

Instrumentation available for profiling Java applications:•Execution Time Analysis•Memory Analysis•Thread Analysis•User written probes



Profiling Options

Data collection can be expensive– Application under test slows down– Large amount of data for both man and

machine

Filter Sets– Collect data of interest– Focus on your code– Can be exported for sharing



Profiling Metrics Determine application behavior

– Execution time, number of calls, call tree

– Memory allocation



Thread metrics Thread Statistics

Threads Visualizer – view timeline of threads with state



Custom Profiling– Implement your own Java probes to collect runtime information from your application

Java code fragments– Class variables– Runtime fields

• thisObject• className• …

– Target filters

Custom profiling



Memory Analysis

Memory Analysis is a profiling analysis type for Java 1.5+ which displays collected dataabout Objects during application execution

Instance data collection is a new option that can be enabled during Profiling. It provides theability to inspect the composition of Objects, including obtaining the live Object values,variable names and instance size during the profile session.

The data that has been collected can be exported to an XML file.



Enable Instance Data Collection



Can also import/export to XML file for comparison



Agenda



Profiling


Tuning the JVM


Questions



IBM Support Assistant

Single access point for finding and downloading support tools for IBM products

Also includes many tools for problem determination, including:– JVM Health Center– Thread and Monitor Dump Analyzer

Access from RAD launch-pad



Diagnostic Data

Data from the JVM -– Javacore – thread information– Heapdumps – heap details– Verbose GC logging – garbage collection data

Data from agents connected to the JVM– Health Center– Rational Agent Controller– More detailed information, higher accuracy: thread, heap details, execution flow– More overhead than Javacore or Heapdumps

Data from the Application Server: PMI



General Strategy

Try to isolate issues using data that’s easily available or obtainable:– Javacores, heapdump files– Use wsadmin to interactively send events to JVM MBean

If problem not resolved with this level of information– Narrow area of investigation and use filtering combined with more detailed data collection agents

There is often more than one tool that could be used

Measure, measure, measure!



Crashes

Problem: The untimely death of an application

View WebSphere server logs

Use Thread and Monitor Dump Analyzer–Javacore file from the application–Identifies where the threads last executed



Problem: App is completely unresponsive

Use Rational Application Developer– Complete hangs - run under debugger and pause after unresponsive point– Hangs/deadlocks

• Thread Analysis for more detailed data• Thread statistics and Threads Visualization contention

Use Thread and Dump Analyzer

Hangs & Deadlocks



Memory Leaks / Excessive Memory Use

Problem:– Slowdown due to increased garbage collection– Reduced number of concurrent sessions

Use Rational Application Developer– Use Memory Analysis– Identify offending object types– Restart analysis, tracking object allocation sites if understanding where they’re being

allocated is needed

Consider reducing session timeout to reduce longevity of session objects



Bottlenecks – Overly Aggressive Locking

Problem: slow responsiveness and throughput of web apps

Potential cause: multithreaded code blocks protect more than they need

Use Rational Application Developer– Thread Analysis for more detailed data– Thread statistics and Threads Visualization to view overlapping contention areas over time– Investigate long lock areas

Demo



Bottlenecks – Inefficient Code Paths

Problem: Slow responsiveness and throughput of web apps

Potential cause: redundant calls or long running code

Use Rational Application Developer– Execution time analysis for detailed metrics– Look for long cumulative times and base times

Demo



Agenda



Profiling


Tuning the JVM


Questions



JVM Tuning – Garbage Collection Strategy and Heap size

Potential Issues:– Pause times – server unable to process requests due to GC pause times– Factors affected GC frequency and size/number of session objects supported

• Garbage collection policies• Heap size parameters

Use Health Center – live recommendations for GC strategy and heap size– Execute a reasonable portion of your application first

Use Garbage Collection and Memory Visualizer– Enable verbose GC output– Provides more detailed tuning recommendations

Demo



Agenda



Profiling


Tuning the JVM


Questions



WebSphere Application Server Tuning

Server configuration can affect the performance of your application

Analysis Tools– Enable logging of key PMI data via admin console

• Open tivoli view – showing various tabs that can be monitored• Start logging activity for scenarios you want to optimize for• Use the performance advisor to produce tuning recommendations• Retry the scenario with configuration changes

Demo



Conclusions Finding problems early in development saves money

Code Coverage from Rational Application Developer running on a server– Verify what get’s touched incrementally– Combine with unit tests or BVT to automate and track

Many tools available to aid in problem determination:– Understanding what the various diagnostic data contains helps guide problem determination strategy for

particular problems• Different tools operate on different input data

– Detailed and accurate data collection incur overhead• Identify focus areas to filter data collection

– WebSphere Application Server contains embedded performance monitoring• Useful Java EE centric performance data• Advisor provides server tuning suggests for you applications



© IBM Corporation 2011. All Rights Reserved.

IBM, the IBM logo, and ibm.com are trademarks orregistered trademarks of International Business MachinesCorp., registered in many jurisdictions worldwide. Otherproduct and service names might be trademarks of IBM orother companies. A current list of IBM trademarks isavailable on the Web at “Copyright and trademarkinformation” at www.ibm.com/legal/copytrade.shtml.

Copyright and Trademarks



Export Collected Object Data The data, such as in the example below, can be exported and used for comparison<?xml version="1.0" encoding="UTF-8" ?>

<heap-instance-data version="1">

…<complex-obj tid="418" parent-tid="429" class="Keystore" package="live.instance.test" size="16">

<object tid="423" parent-tid="418" class="Entry:entries" package="" size="24"/>

</complex-obj>

<complex-obj tid="422" parent-tid="423" class="Entry:next" package="" size="24">

<primitive type="String" name="value">Value 1</primitive>

<primitive type="String" name="key">Key 1</primitive>

</complex-obj>

<complex-obj tid="423" parent-tid="418" class="Entry:entries" package="" size="24">

<object tid="422" parent-tid="423" class="Entry:next" package="" size="24"/>

<primitive type="String" name="value">Value 2</primitive>

<primitive type="String" name="key">Key 2</primitive>

</complex-obj>

…</heap-instance-data>

Problem Determination of your WebSphere …...Allow users to create, enable and disable validation...

Documents

Transcript of Problem Determination of your WebSphere …...Allow users to create, enable and disable validation...